A scaler learns its smallest and largest values from the training readings. New readings are then measured against those numbers, even when they fall outside them.
Task: write scale_new(train_values, new_values).
MinMaxScaler on train_values only (one column).new_values, without learning anything from them.